Motor Vehicle Collisions - Crashes Analysis¶
Introduction to Dataset¶
The Motor Vehicle Collisions crash table contains details on the crash event. Each row represents a crash event. The Motor Vehicle Collisions data tables contain information from all police reported motor vehicle collisions in NYC. The police report (MV104-AN) is required to be filled out for collisions where someone is injured or killed, or where there is at least $1000 worth of damage (https://www.nhtsa.gov/sites/nhtsa.dot.gov/files/documents/ny_overlay_mv-104an_rev05_2004.pdf). It should be noted that the data is preliminary and subject to change when the MV-104AN forms are amended based on revised crash details.For the most accurate, up to date statistics on traffic fatalities, please refer to the NYPD Motor Vehicle Collisions page (updated weekly) or Vision Zero View (updated monthly).
Today's Outline¶
 |
|---|
| 6 Step Machine Learning Modelling Framework |
More specifically, we'll look at the following topics.
- Exploratory data analysis (EDA) - the process of going through a dataset and finding out more about it.
- Feature Analysis - since we're predicting the presence of heart disease, are there some things which are more important for prediction?
- Reporting what we've found - if we had to present our work, what would we show someone?
To work through these topics, we'll use pandas, Matplotlib and NumPy for data anaylsis, as well as, Scikit-Learn for machine learning and modelling tasks.
 |
|---|
| Tools which can be used for each step of the machine learning modelling process. |
1. Problem Definition¶
There really is no particular problem we're trying to solve. We're just analyzing data and looking for insights that we can use to draw conclusions.
There really is no particular problem we're trying to solve. We're just analyzing data and looking for insights that we can use to draw conclusions.
Questions¶
- Total Number of People Killed per Borough?
- Number of People Injured VS Killed per Month?
- Number of Accidents per Month?
- Which Borough has the highest number of deaths by collisions?
- What Are The Top 5 Causes of Collisions In Brooklyn, Queens & Manhattan?
- What is the weekly collisions spread over every hour?
- What is the collisions frequency every hour of the week?
- How do collisions that cause injuries vary by street ?
Import Necessary Tools¶
In [139]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import chart_studio.plotly as py
import plotly.graph_objs as go
from plotly.offline import init_notebook_mode, iplot
init_notebook_mode()
import plotly.offline as py
py.init_notebook_mode(connected=True)
import plotly.graph_objs as go
import plotly.tools as tls
import warnings
warnings.filterwarnings('ignore')
import time
import networkx as nx
from plotly.graph_objs import *
2. Data¶
What you'll want to do here is dive into the data your problem definition is based on. This may involve, sourcing, defining different parameters, talking to experts about it and finding out what you should expect.
We'll look for Attributes (also called features) and variables that we'll potentially use to predict our target variable.
Attributes and features are also referred to as independent variables and a target variable can be referred to as a dependent variable.
Load Data¶
In [160]:
df = pd.read_csv('Datasets/Motor_Vehicle_Collisions_-_Crashes.csv')
collisionsDF = df.copy()
df_copy = df.copy()
df_copy3 = df.copy()
all_dataC = df.copy()
df.head().T
Out[160]:
EDA (Exploratory Data Analysis)¶
In [143]:
df.shape
df.describe()
df.info()
Out[143]:
In [48]:
df.columns
Out[48]:
In [49]:
#Number of accidents per Borough
df.groupby('BOROUGH').size()
Out[49]:
In [51]:
df['ZIP CODE'].isna().sum()
Out[51]:
In [142]:
# check missing values
df.isnull().any()
Out[142]:
Back To Questions That Sparked My Interests¶
- Total Number of People Killed per Borough
- Number of People Injured / Killed Per Month?
- Number of People Killed Per Borough
- Number of People Killed Per Month
- What Are the Number of Accidents Per Day
- What is the Hourly Rate of Deaths
- How is the weekly collisions spread over every hour?
- What is the collisions frequency every hour of the week?
- How do collisions that cause injuries vary by street ?
2.1 What is the total number of people killed by borough?¶
In [52]:
print(df[["BOROUGH","NUMBER OF PERSONS INJURED"]].groupby("BOROUGH").count())
In [55]:
#NUMBER OF PERSONS Killed per BOROUGH
print(df[["BOROUGH","NUMBER OF PERSONS KILLED"]].groupby("BOROUGH").count())
print(df[["BOROUGH","NUMBER OF PEDESTRIANS KILLED"]].groupby("BOROUGH").count())
In [56]:
#getting the counts of people killed by borough then rearranging the columns
persk_data=df[["BOROUGH","NUMBER OF PERSONS KILLED"]].groupby("BOROUGH").count()
persk_data['BOROUGH'] = persk_data.index
persk_data=persk_data.reset_index(drop=True)
pedk_data=df[["BOROUGH","NUMBER OF PEDESTRIANS KILLED"]].groupby("BOROUGH").count()
pedk_data['BOROUGH'] = pedk_data.index
pedk_data=pedk_data.reset_index(drop=True)
motk_data=df[["BOROUGH","NUMBER OF MOTORIST KILLED"]].groupby("BOROUGH").count()
motk_data['BOROUGH'] = motk_data.index
motk_data=motk_data.reset_index(drop=True)
In [141]:
s1 = persk_data.set_index('BOROUGH')['NUMBER OF PERSONS KILLED']
s2 = pedk_data.set_index('BOROUGH')['NUMBER OF PEDESTRIANS KILLED']
s3 = motk_data.set_index('BOROUGH')['NUMBER OF MOTORIST KILLED']
df1 = (s1+s2+s3).reset_index(name='Total People Killed')
print (df1)
In [144]:
import matplotlib.pyplot as plt
plt.rcdefaults()
fig, ax = plt.subplots()
y_pos= np.arange(df1.shape[0])
y=df1["Total People Killed"].values
names=df1["BOROUGH"].values
ax.barh(y_pos, y, align='center',color='green', ecolor='black')
ax.set_yticks(y_pos)
ax.set_yticklabels(names)
ax.invert_yaxis() # labels read top-to-bottom
ax.set_xlabel('People Killed in Accident')
ax.set_title('Number of People Killed Per Borough')
for i, v in enumerate(y):
ax.text(v + 3, i + .25, str(v), color='Green', fontweight='bold')
plt.show()
2.2 What is the number of people injured vs killed per month?¶
In [22]:
df_copy['DATE'] = pd.to_datetime(df_copy['CRASH DATE'], format="%m/%d/%Y")
day_data = df_copy\
.filter(items=['DATE','NUMBER OF PERSONS INJURED','NUMBER OF PERSONS KILLED'])\
.groupby([df_copy.DATE.dt.year, df_copy.DATE.dt.month, df_copy.DATE.dt.day])\
.sum()
day_data['date'] = day_data.index
day_data['date'] = pd.to_datetime(day_data['date'], format="(%Y, %m, %d)")
day_data = day_data.reset_index(drop=True)
In [23]:
day_data=day_data.tail(100)
In [24]:
# # import plotly
# import plotly.plotly as py
# import plotly.graph_objs as go
# # these two lines are what allow your code to show up in a notebook!
from plotly.offline import init_notebook_mode, iplot
init_notebook_mode()
# making a chart
trace1 = go.Scatter(
x=day_data.date, # assign x as the dataframe column 'x'
y=day_data['NUMBER OF PERSONS INJURED'],
name='INJURED'
)
trace2 = go.Scatter(
x=day_data.date,
y=day_data['NUMBER OF PERSONS KILLED'],
name='KILLED'
)
plotted_data = [trace1, trace2]
layout = dict(title="Number of persons injured/killed per month",
xaxis=dict(title='Date', ticklen=10, zeroline=False))
fig = dict(data=plotted_data, layout=layout)
iplot(fig)
2.3 Number of Accidents per Month¶
In [32]:
#delete empty rows
df=df.dropna(how='all')
#reset index to remove missing indexes
df=df.reset_index(drop=True)
#split date from "DATE" column
df['Date']=df['CRASH DATE'].apply(lambda x: x[:10])
In [33]:
# parse dates
for i in range(len(df['Date'])):
if(df['Date'][i].startswith('00')):
p=df['Date'][i-1].split('-')
current=df['Date'][i].split('-')
current[0]=p[0]
df['Date'][i]="-".join(current)
df['Date'] = pd.to_datetime(df['Date'], format = "%m/%d/%Y")
df['Date'][:5]
Out[33]:
In [34]:
# count of accidents per month
acc_per_month = df['Date'].groupby([df.Date.dt.year, df.Date.dt.month]).agg('count')
# convert to dataframe
acc_per_month = acc_per_month.to_frame()
# move date month from index to column
acc_per_month['date'] = acc_per_month.index
# rename column
acc_per_month = acc_per_month.rename(columns={acc_per_month.columns[0]:"acc"})
# re-parse dates
acc_per_month['date'] = pd.to_datetime(acc_per_month['date'], format="(%Y, %m)")
# remove index
acc_per_month = acc_per_month.reset_index(drop=True)
# get month of meet
acc_per_month['month'] = acc_per_month.date.dt.month
In [154]:
# these two lines are what allow code to show up in a notebook!
from plotly.offline import init_notebook_mode, iplot
init_notebook_mode()
# date on the x axis and accidents on the y axis
dataPlot2 = [go.Scatter(x=acc_per_month.date, y=acc_per_month.acc)]
# the layout of figure
layout = dict(title = "Number of Accidents per Month",
xaxis= dict(title= 'Accidents',ticklen= 2,zeroline= False))
# create and show figure
fig = dict(data = dataPlot2, layout = layout)
iplot(fig)
In [153]:
# def getAccidentsAndPlot(borough):
# dfs = pd.read_csv('Datasets/Motor_Vehicle_Collisions_-_Crashes.csv')
# acc_per_day = pd.DataFrame()
# #Number of accidents per day per in Brooklyn
# dfs = dfs[dfs["BOROUGH"]==borough]
# # count of accidents per day in Brooklyn
# acc_per_day = dfs['CRASH DATE'].groupby([dfs['CRASH DATE'].dt.year, dfs['CRASH DATE'].dt.month,dfs['CRASH DATE'].dt.day]).agg('count')
# # convert to dataframe
# acc_per_day = acc_per_day.to_frame()
# # move date day from index to column
# acc_per_day['date'] = acc_per_day.index
# # rename column
# acc_per_day = acc_per_day.rename(columns={acc_per_day.columns[0]:"acc"})
# # re-parse dates
# acc_per_day['date'] = pd.to_datetime(acc_per_day['date'], format="(%Y, %m, %d)")
# # remove index
# acc_per_day = acc_per_day.reset_index(drop=True)
# # get day of meet
# acc_per_day['day'] = acc_per_day.date.dt.day
# # date on the x axis and accidents on the y axis
# dataPlot3 = [go.Scatter(x=acc_per_month.date, y=acc_per_month.acc)]
# # the layout of figure
# layout = dict(title = "Number of Accidents per Day",
# xaxis= dict(title= borough,ticklen= 2,zeroline= False))
# # create and show figure
# fig = dict(data = dataPlot3, layout = layout)
# iplot(fig)
2.4 Which Borough has the highest number of deaths by collisions?¶
In [155]:
# these two lines are what allow code to show up in a notebook!
from plotly.offline import init_notebook_mode, iplot
init_notebook_mode()
dataPlot = [go.Histogram(histfunc = "sum",x=names,y=y)]
# the layout of figure
layout = dict(title = "Number of Deaths Per Borough",
xaxis= dict(title= 'Borough',ticklen= 2,zeroline= False))
# create and show figure
fig = dict(data = dataPlot, layout = layout)
iplot(fig)
Let's Focus On Brooklyn, Since it has the largest number of deaths in New York
2.5 What Are The Top 5 Causes of Collisions In Brooklyn?¶
In [156]:
#Number of Accidents Pe Day in Queens
# getAccidentsAndPlot("QUEENS")
In [62]:
all_dataC.head(3)
Out[62]:
In [157]:
all_dataC.isnull().any()
Out[157]:
Data Cleaning¶
In [76]:
#dtypes
all_dataC['CRASH TIME'] = pd.to_datetime(all_dataC['CRASH TIME'])
all_dataC['ZIP CODE'] = pd.to_numeric(all_dataC['ZIP CODE'], errors='coerce')
#rename
all_dataC = all_dataC.rename(columns={'CONTRIBUTING FACTOR VEHICLE 1': 'collision_factor'})
#add count column
all_dataC['count']=1
#extract datetime data to create 3 additional columns
all_dataC['year'] = all_dataC['CRASH TIME'].dt.year
#not null or unspecified
all_dataC = all_dataC[pd.notnull(all_dataC['collision_factor'])]
all_dataC = all_dataC.drop(all_dataC[all_dataC['collision_factor']=='Unspecified'].index)
all_dataC['year'] = all_dataC['CRASH TIME'].dt.year
#select columns
all_dataC = all_dataC[['year', 'BOROUGH', 'collision_factor', 'COLLISION_ID', 'NUMBER OF PERSONS INJURED']]
In [112]:
all_dataC.head()
Out[112]:
Brooklyn¶
In [133]:
#top collision factors by year, by borough
df_agg = all_dataC.groupby(['BOROUGH','year', 'collision_factor']).sum().reset_index()
# g = df_agg['count'].groupby(level=0, group_keys=False)
brooklyn = df_agg[df_agg['BOROUGH']=='BROOKLYN']
brooklyn.sort_values(by=['NUMBER OF PERSONS INJURED'], ascending=False).head()
Out[133]:
Queens¶
In [137]:
#top collision factors by borough
#top collision factors by year, by borough
df_agg = all_dataC.groupby(['BOROUGH','year', 'collision_factor']).sum().reset_index()
# g = df_agg['count'].groupby(level=0, group_keys=False)
brooklyn = df_agg[df_agg['BOROUGH']=='']
brooklyn.sort_values(by=['NUMBER OF PERSONS INJURED'], ascending=False).head()
Out[137]:
Manhattan¶
In [138]:
#top collision factors by borough
#top collision factors by year, by borough
df_agg = all_dataC.groupby(['BOROUGH','year', 'collision_factor']).sum().reset_index()
# g = df_agg['count'].groupby(level=0, group_keys=False)
brooklyn = df_agg[df_agg['BOROUGH']=='MANHATTAN']
brooklyn.sort_values(by=['NUMBER OF PERSONS INJURED'], ascending=False).head()
Out[138]:
2.6 What is the weekly collisions spread over every hour?¶
In [37]:
collisionsDF.head(2)
Out[37]:
In [38]:
# Assuming that dataframe has DATE and Time columns sorted (meaning row 0 is latest date and last row represents the earliest date)
# It can be suppossed based on the tail() and head() results
oldest_date = collisionsDF.iloc[-1]['CRASH DATE'] #tail
newest_date = collisionsDF.iloc[0]['CRASH DATE'] #head
newest_date
Out[38]:
In [39]:
#Extract DateTime features
collisionsDF['collisions_datetime'] = pd.to_datetime(collisionsDF['CRASH DATE'], infer_datetime_format = True) #infer_datetime_format=True)
collisionsDF['collisions_time'] = pd.to_datetime(collisionsDF['CRASH TIME'], infer_datetime_format = True)
#collisionsDF.loc[:, 'collisions_datetime'] = collisionsDF['collisions_datetime'].dt.date
collisionsDF.loc[:, 'collisions_weekday'] = collisionsDF['collisions_datetime'].dt.weekday
collisionsDF.loc[:, 'collisions_weekofyear'] = collisionsDF['collisions_datetime'].dt.weekofyear
collisionsDF.loc[:, 'collisions_hour'] = collisionsDF['collisions_time'].dt.hour
collisionsDF.loc[:, 'collisions_minute'] = collisionsDF['collisions_time'].dt.minute
collisionsDF.loc[:, 'collisions_dt'] = (collisionsDF['collisions_datetime'] - collisionsDF['collisions_datetime'].min()).dt.total_seconds()
#collisionsDF['collisions_time'].dtype
collisionsDF.head()
Out[39]:
In [41]:
#Data Preparation for the weekly analysis
#get count of collisions every hour on every weekday.
sunday = collisionsDF[collisionsDF['collisions_weekday'] == 6]
df_sundayhourlytripcount = sunday.groupby('collisions_hour').count()
monday = collisionsDF[collisionsDF['collisions_weekday'] == 0]
df_mondayhourlytripcount = monday.groupby('collisions_hour').count()
tuesday = collisionsDF[collisionsDF['collisions_weekday'] == 1]
df_tuesdayhourlytripcount = tuesday.groupby('collisions_hour').count()
wednesday = collisionsDF[collisionsDF['collisions_weekday'] == 2]
df_wednesdayhourlytripcount = wednesday.groupby('collisions_hour').count()
thursday = collisionsDF[ collisionsDF['collisions_weekday'] == 3]
df_thursdayhourlytripcount = thursday.groupby('collisions_hour').count()
friday = collisionsDF[collisionsDF['collisions_weekday'] == 4]
df_fridayhourlytripcount = friday.groupby('collisions_hour').count()
saturday = collisionsDF[collisionsDF['collisions_weekday'] == 5]
df_saturdayhourlytripcount = saturday.groupby('collisions_hour').count()
collisions_hr_x = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23]
sun_crashcounty = df_sundayhourlytripcount['CRASH TIME']
mon_crashcounty = df_mondayhourlytripcount['CRASH TIME']
tues_crashcounty = df_tuesdayhourlytripcount['CRASH TIME']
wed_crashcounty = df_wednesdayhourlytripcount['CRASH TIME']
thurs_crashcounty = df_thursdayhourlytripcount['CRASH TIME']
fri_crashcounty = df_fridayhourlytripcount['CRASH TIME']
sat_crashcounty = df_saturdayhourlytripcount['CRASH TIME']
df_sundayhourlytripcount.tail(1)
sun_crashcounty.tail(1)
# Create traces
trace1 = go.Scatter(x = collisions_hr_x, y = sun_crashcounty, mode = "lines+markers", name = 'Sunday')
trace2 = go.Scatter(x = collisions_hr_x, y = mon_crashcounty, name = 'Monday')
trace3 = go.Scatter( x = collisions_hr_x, y = tues_crashcounty, name = 'Tuesday')
trace4 = go.Scatter( x = collisions_hr_x, y = wed_crashcounty, name = 'Wednesday')
trace5 = go.Scatter( x = collisions_hr_x, y = thurs_crashcounty, name = 'Thursday')
trace6 = go.Scatter( x = collisions_hr_x, y = fri_crashcounty, name = 'Friday')
trace7 = go.Scatter(x =collisions_hr_x, y = sat_crashcounty, mode = "lines+markers", name = 'Saturday')
layout = dict(title = 'Weekly Collisions by Hour',
xaxis= dict(title= 'Hour of the day',ticklen= 5,zeroline= False),
yaxis= dict(title= 'Collisions quantity',ticklen= 5,zeroline= False)
)
linedata = [trace1, trace2, trace3, trace4, trace5, trace6,trace7]
fig = dict(data=linedata, layout=layout)
py.iplot(fig, filename='timeline-lineplot')
2.7 What is the collisions frequency every hour of the week?¶
In [65]:
#Preparing data for the graph
W_0 = collisionsDF[collisionsDF['collisions_weekday'] == 0].groupby('collisions_hour').count()
W_1 = collisionsDF[collisionsDF['collisions_weekday'] == 1].groupby('collisions_hour').count()
W_2 = collisionsDF[collisionsDF['collisions_weekday'] == 2].groupby('collisions_hour').count()
W_3 = collisionsDF[collisionsDF['collisions_weekday'] == 3].groupby('collisions_hour').count()
W_4 = collisionsDF[collisionsDF['collisions_weekday'] == 4].groupby('collisions_hour').count()
W_5 = collisionsDF[collisionsDF['collisions_weekday'] == 5].groupby('collisions_hour').count()
W_6 = collisionsDF[collisionsDF['collisions_weekday'] == 6].groupby('collisions_hour').count()
#W_0.head()
trace = go.Heatmap(z=[W_0['CRASH TIME'],W_1['CRASH TIME'],W_2['CRASH TIME'],W_3['CRASH TIME'],W_4['CRASH TIME'],W_5['CRASH TIME'],W_6['CRASH TIME']],
y=['Monday','Tuesday','Wednesday','Thursday','Friday','Saturday', 'Sunday'],
x=['Midnight','1am','2am','3am','4am','5am','6am','7am','8am','9am','10am','11am','Noon','1pm','2pm',
'3pm','4pm','5pm','6pm','7pm','8pm','9pm','10pm','11pm'],
colorscale='Reds',xgap = 10,ygap = 10)
layout = dict(title = 'Collisions per week per hour')
dataheat=[trace]
fig = dict(data = dataheat, layout=layout)
py.iplot(fig, filename='labelled-heatmap')
#Takeaway: Friday seems to have rush hours around 5pm. This may be impacting durations trip around that time. Could be interesing to know what
#is the street name with most collisions in that hour.
2.8 How do collisions that cause injuries vary by street ?¶
In [71]:
collisionsDF.groupby('ON STREET NAME').count().sort_values(by='NUMBER OF PERSONS INJURED', ascending=False).head(10)['NUMBER OF PERSONS INJURED']
Out[71]:
In [72]:
TOP5 = collisionsDF.groupby('ON STREET NAME').count().sort_values(by='NUMBER OF PERSONS INJURED', ascending=False).head(5)['NUMBER OF PERSONS INJURED']
TOP5.keys()
Out[72]:
In [87]:
TOP5.keys()[4]
Out[87]:
In [77]:
#Lets look at demand per hour.
bar0data = collisionsDF[collisionsDF['ON STREET NAME']==TOP5.keys()[0]].groupby(['collisions_hour']).count().reset_index()
bar1data = collisionsDF[collisionsDF['ON STREET NAME']==TOP5.keys()[1]].groupby(['collisions_hour']).count().reset_index()
bar2data = collisionsDF[collisionsDF['ON STREET NAME']==TOP5.keys()[2]].groupby(['collisions_hour']).count().reset_index()
bar3data = collisionsDF[collisionsDF['ON STREET NAME']==TOP5.keys()[3]].groupby(['collisions_hour']).count().reset_index()
bar4data = collisionsDF[collisionsDF['ON STREET NAME']==TOP5.keys()[4]].groupby(['collisions_hour']).count().reset_index()
In [97]:
trace0 = go.Bar(
x=['Midnight','1am','2am','3am','4am','5am','6am','7am','8am','9am','10am','11am','Noon','1pm','2pm',
'3pm','4pm','5pm','6pm','7pm','8pm','9pm','10pm','11pm'],
y= bar0data['NUMBER OF PERSONS INJURED'],
name='BROADWAY',
marker=dict(color='rgb(49,130,189)' ))
trace1 = go.Bar(
x=['Midnight','1am','2am','3am','4am','5am','6am','7am','8am','9am','10am','11am','Noon','1pm','2pm',
'3pm','4pm','5pm','6pm','7pm','8pm','9pm','10pm','11pm'],
y= bar1data['NUMBER OF PERSONS INJURED'],
name='ATLANTIC AVENUE',
marker=dict(color='rgb(255,192,203)',))
trace2 = go.Bar(
x=['Midnight','1am','2am','3am','4am','5am','6am','7am','8am','9am','10am','11am','Noon','1pm','2pm',
'3pm','4pm','5pm','6pm','7pm','8pm','9pm','10pm','11pm'],
y= bar2data['NUMBER OF PERSONS INJURED'],
name='BELT PARKWAY',
marker=dict(color='rgb(138,43,226)',))
trace3 = go.Bar(
x=['Midnight','1am','2am','3am','4am','5am','6am','7am','8am','9am','10am','11am','Noon','1pm','2pm',
'3pm','4pm','5pm','6pm','7pm','8pm','9pm','10pm','11pm'],
y= bar3data['NUMBER OF PERSONS INJURED'],
name='3 Avenue',
marker=dict(color='rgb(255,215,0)',))
trace4 = go.Bar(
x=['Midnight','1am','2am','3am','4am','5am','6am','7am','8am','9am','10am','11am','Noon','1pm','2pm',
'3pm','4pm','5pm','6pm','7pm','8pm','9pm','10pm','11pm'],
y= bar4data['NUMBER OF PERSONS INJURED'],
name='NORTHERN BOULEVARD',
marker=dict(color='rgb(255,0,0)',))
datax = [trace0, trace1, trace2, trace3, trace4]
layout = go.Layout(xaxis=dict(tickangle=-45),barmode='group')
fig = go.Figure(data=datax, layout=layout)
py.iplot(fig)
The rush hour at Broadway, Atlantic avenue and 3 avenue is 4 pm while at Northern Boulevard is 5pm. However, between 4pm and 6pm is the range of time with the most hight risk of collision.
Conclusions¶
Proposal¶
Since driver inattention and distraction is the highest collision factor Y/Y and most accidents occur during workweeks and business hours, I propose a data driven based public awareness campaign focused on reducing driver distractions.
Real World Initiatives¶
In April 9th, 2019 the Govenor's Trafic Saftey Committee (GTSC) launched a Distracted Driving PSA Campaign in NYC. The campaign featured on billboards, television, and radio.
My Reccommendation¶
- I reccommend airing the campaign during peak commute hours and during holidays during COIVID 19.
- Using social media in the campaign strategy. We can start by using zip codes with highest death rates
- Geospatial social media analysis
- Eat a snickers after work
- I believe this campaign will have an impact on the number of collisions and that it will see a reduction in the number of injuries from reckless driving.
Future Improvements¶
- Implement machine learning: time series analysis
- 3D Visualizations
- Heatmap
- Confidence Intervals
- Hypothesis Testing